Why are the optimization variable and penalty term in the objective function always of the same order of magnitude? (No matter how I adjust rho)

1 
cvx_quiet(false);
cvx_begin
    variable W_k(K, K) complex
    variable Q_k(K*L,K) complex
    variable gamma_k nonnegative
    
    expression penalty
    expression y(K,L)

    penalty=pow_pos(norm(Q_k-(U.')*big_sigma*W_k+rho*lambda_q),2);

    for k=1:K
        for l=1:L
            y(k,l)=2*real(conj(mu_opt(k,l))*Q_k((k-1)*L+l,k))-abs(mu_opt(k,l))^2*sigma;
            for i=1:K
                if i~=k
                    y(k,l)=y(k,l)-abs(mu_opt(k,l))^2*pow_pos(abs(Q_k((k-1)*L+l,i)),2);
                end
            end
        end
    end

    maximize gamma_k-1/(2*rho)*penalty

    subject to
    min(min(y))>=gamma_k;
    sum_square_abs(W_k(:,1)) + sum_square_abs(W_k(:,2)) + sum_square_abs(W_k(:,3)) + sum_square_abs(W_k(:,4))  <= P_max;

    cvx_end
2

‘gamma_k’ and ‘1/(2*rho)*penalty’ are always of the same order of magnitude